test(adapter): comprehensive HomegrownAdapter unit tests

[Jesssullivan]

Summary

• Expands HomegrownAdapter test coverage from 3 superficial constructor checks to 28 behavioral tests
• Mocks Drizzle DB layer with fluent chain builder to test service resolution (UUID vs acuityId), provider lookup (solo practice pattern), client find-or-create, reservation create/release, booking cancellation, and Effect error wrapping
• Null→undefined mapping, empty result handling, and non-Error throw wrapping validated

Coverage areas

• Services: getServices (4 tests), getService (3 tests)
• Providers: getProviders (3), getProvider (2), getProvidersForService (1)
• Reservations: createReservation (2), releaseReservation (1)
• Clients: findOrCreateClient (2), getClientByEmail (3)
• Bookings: cancelBooking (2)
• Effect wrapping: DB errors (2)
• Config: creation + method existence + custom config (3)

Availability math (getAvailableDates, getAvailableSlots, checkSlotAvailability) delegates to availability-engine.ts which already has 39 dedicated tests.

Test plan

• All 28 new tests pass
• Full suite: 606 tests pass (21 files), 0 failures
• No type errors

Tracker: TIN-104 (package quality)

[greptile-apps]

Greptile Summary

This PR replaces 3 superficial constructor checks with 28 behavioral unit tests for HomegrownAdapter, using a fluent mock-DB chain to cover service/provider lookup, client find-or-create, reservation lifecycle, booking cancellation, and Effect error wrapping. All findings are P2.

• The four most complex methods — createBooking, createBookingWithPaymentRef, getBooking, and rescheduleBooking — have no unit tests; createBooking in particular nests a live Effect.runPromise call that is worth exercising.
• The isUuid routing branch in resolveService is not validated because the where mock discards its argument.
• beforeEach is imported from vitest but never used.

Confidence Score: 5/5

Test-only PR; all findings are P2 suggestions and do not block merge.

No production code is changed. The 28 new tests are correctly structured, the mock chain faithfully mirrors the Drizzle fluent API, and all assertions align with the implementation. The only gaps (missing booking-method tests, unvalidated UUID branch, unused import) are style/coverage suggestions that don't indicate correctness problems in the tests that do exist.

No files require special attention.

Important Files Changed

Filename	Overview
src/adapters/tests/homegrown-adapter.test.ts	Expands HomegrownAdapter unit tests from 3 constructor checks to 28 behavioral tests using a fluent mock-DB chain; findings are P2 only (unused beforeEach import, no coverage for createBooking/getBooking/rescheduleBooking, UUID routing branch not validated).

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[createMockDb] --> B[chainTerminals\nlimit · orderBy · returning]
    A --> C[chain\nfrom · where · values · set]
    C --> D[from returns\nwhere · orderBy · limit]
    D -->|.where| B

    subgraph getServices
        E[select · from · where · orderBy] --> B
    end

    subgraph getService / getProvider / getClientByEmail
        F[select · from · where · limit] --> B
    end

    subgraph findOrCreateClient
        G[select · where · limit] --> B
        H[insert · values · returning] --> B
        I[update · set · where] -->|resolves undefined| J[void]
    end

    subgraph cancelBooking / releaseReservation
        K[update · set · where] -->|resolves undefined| J
    end

    subgraph NOT TESTED
        L[createBooking\nfindOrCreateClient + resolveService + insert]
        M[getBooking\n4x select joins]
        N[rescheduleBooking\nselect + update + getBooking]
    end

Loading

_{Reviews (1): Last reviewed commit: "test(adapter): comprehensive HomegrownAd..." | Re-trigger Greptile}

[greptile-apps]

Unused import: beforeEach

beforeEach is imported from vitest but never used in the test file. This will trigger a lint/TS unused-import warning.

Suggested change

	// Mock schemas — minimal shape matching what Drizzle ORM tables expose
	import { describe, it, expect, vi } from 'vitest';

[greptile-apps]

Complex booking methods have no unit tests

createBooking, createBookingWithPaymentRef, getBooking, and rescheduleBooking — the four most structurally complex methods in the adapter — are absent from this suite. createBooking is particularly notable: it nests a live Effect.runPromise(adapter.findOrCreateClient(...)) call inside fromAsync, making it the only method that can surface an unexpected error from a nested Effect if the composition assumption ever changes. getBooking performs four separate select calls and reconstructs the full Booking shape including a conditional practitioner lookup.

Worth adding these tests to complete the scope, even with the same fluent-chain mock pattern used elsewhere.

[greptile-apps]

UUID routing condition not verified

getService with a full UUID and with a numeric acuityId take different code paths through the isUuid branch in resolveService — a UUID triggers or(eq(id,...), eq(acuityId,...)) while a non-UUID uses only eq(acuityId,...). Because the mock ignores the where argument, both tests resolve identically from the canned row and the branch logic is never exercised. Inspecting the where spy's call argument would let you assert that the UUID path actually builds the compound or(...) condition.